Transcription Guide

Introduction
Transcription is the first step of our 4-step process. As a certified transcriber on Scribie.com you
can log in to your account and select any of the available files for transcription. After you submit
the assignment, it will go through the rest of the 4-steps of our process and delivered to the
customer. Your submission will also be graded based on the changes and the earning for the file
will be credited to your account based on the grade. Your earnings will accumulate in your
account and you can withdraw it to your PayPal account anytime.
This guide provides the rules and other details about the raw transcription step of the process. It
is intended for our certified transcribers and also as a reference. For any questions or
suggestions, please contact support.

Payment Rate
Aim for QC promotion for higher payment rates
The payment rate for transcription varies from $5/ah to $25/ah (ah stands for audio hour). We
also pay a monthly bonus for 3 or more hours completed each month. The rates are higher for
later steps in our process as there are fewer people there. Therefore QCs are paid the highest
rates.
The rates are determined by an algorithm which takes into account various factors such as
deadline, issues in the audio file, load on our system, etc. If you see only lower rated files, that's
because all the higher rated files have already been selected.
The payment is credited once the file has been delivered to the customer and graded. Rejected
files are not paid.

Transcription Guidelines
The transcription guideline is a set of rules that should be always followed. Please memorize
these rules and adhere to them as strictly as possible. All these rules are considered while
grading.
Never violate these rules
Files which do not contain any spoken audio and/or contain only background conversation
should be reported instead of being submitted with a placeholder. Please also report any files
with poor audio quality, distortions, distant speakers, high background or ambient noise and/or
difficult accents so that it can be investigated. Such files may be returned to the customer which
will lead to your submission being automatically canceled.

I. The raw transcript should not contain text other than the spoken audio. No
headers/footers, speaker tracking, time-codes, comments, etcetera.
 II. No part of the audio should be omitted, unless specified otherwise by an instruction or
another guideline.
III. Inaudible parts should be omitted and marked with a blank ____.
IV. Laughter should be omitted and marked with [laughter] or [chuckle].
V. Ellipsis ... should be used to indicate unfinished sentences or mid-sentence pauses.
VI. Contractions, wanna, gonna, kinda, gotta, should not be expanded.
VII. Numbers one to nine should be transcribed in letters and numerals otherwise, except for
measurements.
VIII. The number of mistakes should not be greater than 2 for every 10 comprehensible words
spoken in the audio.
IX. A new paragraph should be started at each change of speaker and paragraphs should be
separated by an empty line.
X. Fillers (right, you know, I think, like, I mean, so, actually, basically), false starts and
stutters should be omitted, unless necessary, and should be included for strict verbatim files.
XI. Utterances should be omitted for non-strict verbatim files, unless necessary. For strict
verbatim files, only the following utterances should be transcribed: Mm-hmm if in
agreement, Uh-huh if in disagreement, Umm and Uh.

Additional Notes:

 The minimum accuracy requirement is 2 major changes per minute of audio. Blanks
which are filled during review are considered mistakes.
 For Guideline VII, measurements are considered as anything which has a unit (kg, ml,
currency units, etcetera).
 For Guideline X, the filler words specified have to be frequently repeated by a speaker
during the conversation to be considered a filler. In-frequent usage of those words should be
transcribed.
 False starts are when the speaker starts a statement but immediately corrects it to
something else.
 For Guideline XI, all files are non-strict verbatim by default. Strict verbatim files are
indicated by an instruction on the file. Therefore if the instruction is not present, all
utterances, fillers, false starts, stutters should be omitted. If the instruction is present,
nothing should be omitted and the raw transcript should match the audio word for word.
 The only case when an utterance is required in a non-strict verbatim file is when it is a
response to a direct question. The unless necessary part of Guideline XI specifies that.

Style Suggestions
The following are some styles that we recommend. Please refer to the Style Guide for a more
comprehensive list.

 Punctuations are free style. Please use your discretion on placement of commas, periods,
semicolons, exclamations, etc.
 Spelling styles are different for each file and the styles suggested by the spell check
should be followed.
  All proper nouns should be first letter capitalized, e.g., names of places, person,
organizations, job titles, etcetera.
 Spelled out names should be all capitalized and separated by hyphens, e.g., A-D-A-M.
 Use etcetera instead of etc.
 Use i.e., instead of ie.
 Use e.g., instead of Eg.
 Alright and All right are both acceptable.
 A statement can begin with an And.
 Generic drug names should be all lowercase and branded drug names should be first letter
capitalized. Google once to check if they are generic or branded.
 Special marking schemes for phonetics and spelling should not be used. Blanks should be
used instead.
 Periods should not be used in acronyms, e.g., PhD instead of Ph.D.
 Phonetics such as Alpha, Bravo, Tango should be transcribed as ABT.
 The following examples further illustrate Transcription Guideline VII.
o Time: 10:00 AM
o Date: March 2, 2012
o Temperature: 5 degrees
o Money: $5, $500,000, $5 Billion
o Percentage: 5%
o Speed: 5 miles per hour
o Dimension: 6 x 10, 5 feet 2 inches
o Address: 213, 2nd Street
o Era: in your 20s, in the '90s
o Rank: ninth, 10th

Sample Raw Transcript

The following is a sample raw transcript which illustrates the formatting required.
Highlights

 There is no speaker tracking, e.g., interviewer/interviewee, person 1/person 2 etc. There

are also no time codes. Both of these are not required for raw transcript.
 The transcript is broken into paragraphs on each change of speaker. Each speaker's
continuous diction is on one single paragraph and paragraphs are separated by empty lines.
 Inaudible parts are marked with a blank, ____ to make it stand out from the rest of the
transcript. These blanks may or may not get filled during review, but are always checked.

Automated Transcript
Correct and submit the automated transcript
We also provide an automated transcript with each file. The automated transcript is machine
generated and the accuracy varies from 60% to 90% depending upon the audio quality, accents,
and other things. Therefore it has to be corrected before submitting. However, this can still save
around 50% of the typing effort. The following has to be corrected:

 Mistakes
 Missing words
 Missing or incorrect paragraph breaks between speaker turns
 Utterances for strict verbatim files
 Incorrect capitalization
 Missing punctuations, especially "

Difficulty Levels
We provide guidance for difficulty level with each file. The difficulty levels are: Low, Medium
and High. These levels indicate the amount of changes required with respect to the automated
transcript to achieve 99% accuracy. Please refer to the following table for the percentage of
words which have to be corrected.

Difficulty Level Corrections Percentage

Low Less than 10% of all words

Medium Around 10-20% of all words

High More than 20% of all words

We have extensively tested this algorithm and these ranges are accurate to +/- 5%. This
algorithm is based on a Linear Regression model which predicts how many words in the
automated transcript are likely to be wrong based on all the files which we delivered in the last 3
months.

Accents
The accents of the speakers in each file is also provided, whenever available. The accents are
coded as follows:

Accent Code Description

NA North American

CA Canadian
Accent Code Description

AU Australian

GB British

IN Indian

AA African-American

AF African

RW Rwandan

GR German

FR French

IT Italian

PL Polish

EU European

SP Spanish

RU Russian

FN Finnish

TK Turkish
Accent Code Description

ID Indonesian

MX Mexican

HP Hispanic

LA Latin American

BR Brazilian

PR Portugese

NL Dutch

ME Middle Eastern

IR Irish

AS Asian

CN Chinese

KO Korean

SG Singaporean

EA East Asian

NZ New Zealand
Accent Code Description

AB Arabic

MY Malaysian

JP Japanese

SE Southeast Asian

SA South African

JM Jamaican

WI West Indian

AG Aboriginal

SC Scottish

NP Nepalese

EG Egyptian

AI Indigenous American

N/A Unsure/Unknown/Not Applicable

NN Other Non-native/Mixed

Performance
We use a 5 point scale for grading each file. It ranges from A+ (Excellent/5) to D (Poor/1). The
average of these grades, called the Average Transcription Grade, is used to monitor the
performance. The following rules apply for the Average Transcription Grade.

 These rules come into effect only after 1 audio hour of submissions has been credited.
 Only reviewed and rejected files are considered in the average grade; canceled or timed
out files are not counted. Rejected files have a score of 0.

 The minimum Average Transcription Grade is 2.5. Your account will be disabled if it
falls below that mark anytime.
 The running average grade has to be maintained over 2.5 for the last audio hour of
submissions and transcription will be disabled for a week if the average grade is lower.
 The minimum Average Transcription Grade required for promotion to reviewer is 3.

Once transcription is disabled, the only way to get it enabled is to dispute the changes. This
minimum grade is fixed and is enforced automatically and we cannot change it.

Player History
We collect player history data in the Editor for all assignments. The data contains information
about which parts of the audio were played and when. This data is submitted along with the file
and is available to the reviewers, proofreaders and QC. The player history is also used for
performance monitoring and during audits in case a file is returned by the customer.
Assignments therefore have to be completed within the editor and cannot be submitted without a
player history.
Player history also is used to detect inactivity. If no audio is played for more than 2 minutes, the
editor will automatically close. Please re-open the editor whenever you're ready to start again.
Assignments may also be flagged for investigation if the full audio is not played.

Grading Criteria
Audio quality, accents, past performance, etcetera, are not grading factors. The only factor is
major changes.
Grades are based on major changes found in the assignment. The following lists all the types of
changes and their category.

Category Type Description

Soundalike Major Words which sound similar but differ in meaning, e.g., allude and elude.

Mishear Major The spoken word does not match the typed word
 Category Type Description

Omission Major Words which were spoken but not typed

Misspell Major A spelling mistake, e.g., typo

Filler Minor Frequently repeated words or phrases, e.g., like

False start Minor A phrase followed by an immediate correction

Style change Minor A change of style which does not alter the meaning, e.g., percent to %

Subjective change Minor Any change which does not alter the meaning

The following table lists the grade corresponding to the number of major changes, for a 6-minute
file.

Grade Number of Major changes

Rejected 20 or more

Poor (D/1) 16 - 19

Bad (C/2) 12 - 15

Average (B/3) 8 - 11

Good (A/4) 4-7

Excellent (A+/5) 0-3

Diffs
Only the changes highlighted in red are considered in grading
Diffs, short for differences, show the changes that were made during the review. The parts in red
denote the removals whereas the parts in green are the additions. The compare link in the History
tab brings up the diff for any file. The original transcript is not changed during the review; the
changes are made on a copy. Only paragraphs which were changed are included in the diffs.
Paragraphs which are omitted are indicated by a .....
The changes are also summarized in a table below the diff box. Most of the times, we also
provide an audio clip which covers the change so that you can check the change against the
audio. You can dispute the changes made by selecting the change (checkbox on the left) and
clicking the Dispute button at the bottom. Disputes are available only after the file has been
delivered to the customer.
The diff summary will help you improve your performance and learn from others. Please be sure
to check it for each assignment.

Disputes
We encourage you to dispute the changes if it does not correspond to the audio or if the change
should not be major. The disputed changes will be investigated by our admins and resolved. To
raise a dispute, please click the compare link in the History tab and click the Dispute button at
the bottom. This option is available only after the file has been delivered to the customer.
The status of the dispute will be shown in the dispute column of the diff summary table. After
resolution, it will list the category of the dispute and the file will be re-graded.
Changes which have already been marked as major by our admins cannot be disputed.

Tab Completion
Press tab to complete the current word
The editor has built-in tab completion functionality. It provides suggestions as you type in a
popup window below the cursor. You can accept the suggestion by pressing the Tab key or close
the popup by pressing the Escape key. It can also complete frequently occurring words in the file,
even without the popup. Press the Tab key to check if any completions are available after typing
the first 3 characters. Meta's can also be completed by typing [ and the first two words of the
meta.
The tab completion feature can be turned off from the dropdown menu next to the timer.

Dictated Punctuations
The following rules apply for punctuations which are dictated by the speaker. Such files are
typically a single speaker file in which they are either dictating a note or a letter which has to be
typed out. All our Transcription Guidelines still apply, except the rule for spelled out names. The
speakers have to be labelled as S1 always.

 Period, comma, exclamation, question mark and dash should be transcribed as symbols.
 Parenthesis, brackets, and braces should be transcribed literally, e.g., open
bracket and close bracket.
 For lists, bullet items and new paragraph, a new paragraph should be started.
 Spelled out names should be transcribed without the dashes.
Background Conversation
Always transcribe background conversation if it is clear and audible
Background conversation is very subjective and different people perceive it differently.
Customers are certain to return the file if they disagree. Therefore it should be used as sparingly
as possible. The thumb rule is: if it is audible and clear, it should be transcribed. If the file is
strict verbatim, all background conversations should be transcribed.
However, there are some clear cases. E.g., a waiter's voice in between an interview being
conducted in a restaurant is definitely background conversation. The audience chatter before a
speech is also background conversation. Another common example is the discussion amongst
off-camera crew members in a video interview or commercial; the direct questions and answers
are important, the discussion about lighting and other things are background conversation.
On the other hand, if a voice goes faint at any time (e.g. muttering/whispering), then it should be
blank rather than a background conversation. If you are unsure then you should transcribe it or
insert a blank and leave it up to the QAs to decide.
If any part of the audio is omitted as background conversation but transcribed in later stages of
our process, then it will be counted as a mistake. Conversely, if any transcribed parts are
removed and marked as background conversation, it is not counted as a mistake. To be on the
safer side, you should always transcribe all parts of the audio if it is clear and audible.

Monthly Bonus
We pay a monthly bonus of $5 for every 3 hours credited over the period of the calendar month.
The following are the rules for the bonus.

 The amount is credited on the 1st day of each calendar month, for the previous calendar
month.
 Only submissions which have been credited (i.e., delivered to the customer) in the
previous calendar month are counted.
 Submissions which are un-credited are carried over and counted in next month's bonus
calculation.
 Our timezone is +5:30 UTC. Therefore calendar month starts at 12:00 AM UTC +5:30 of
the first day of the month and ends at 12:00 AM UTC +5:30 of the last day of the month.
 You will be notified via email if your Earnings Credit mail notification setting is enabled.

Timeouts
Each transcription assignment has a timeout of 2 hours. The timeout can be extended once after
the initial 2 hours has elapsed. The extension granted is 1 hour. Therefore in total, the assignment
has to be submitted within 3 hours.
The Request Extension option is available in the dropdown menu next to the timer. It is enabled
only when the initial time has elapsed, i.e., when the timer starts blinking.
A grace period between 1-5 minutes is also granted once the timer starts blinking. The timeout
may occur anytime within those 5 minutes.
Once a file has timed out, it can be re-assigned from the History tab and re-submitted within the
grace period, provided it is still available.

Online Editor
We provide an online editor where the transcription has to be done. The Editor contains an audio
player and a text editor where the transcript can be typed. The audio player can be controlled by
buttons or keyboard shortcuts. The text editor is a fully featured text editor (similar to Notepad or
TextEdit). To transcribe, play the file for a few seconds and then pause. Then type everything in
the text box provided. Repeat the process until all the audio has been transcribed. Leave a
blank ____ for portions where the audio is unclear and cannot be understood.
The most used shortcuts are Ctrl+P for toggle play (i.e., pause if audio is playing and play if
audio is paused) and Ctrl+O for skip backwards or rewind. Memorize these shortcuts to use the
Editor effectively. The shortcuts are also customizable. The option is present in the dropdown
menu next to the timer in the Editor. The customizations are saved and can also be reverted back
to default settings later on if desired. We recommend using the latest version of Google Chrome
for the Editor as it has some advanced audio functionalities. Any other HTML5 browser (eg.
Mozilla Firefox) can also be used.
On Google Chrome browser, dictation is also supported. You can dictate instead of typing. It will
save you around 60% of typing effort. Play the file for 5 to 10 seconds, repeat whatever you hear
and then edit out any mistakes. The audio can also be amplified on Google Chrome, from the
dropdown menu next to the timer. It is very useful for files with low volume.

Footpedal Setup Guide

A footpedal can also be used with the Editor by configuring it to emulate a key combination for
each pedal button. That key combination should be one of the shortcuts of the Editor so that the
desired action is triggered. The play, pause and skip backward actions are mostly mapped to
pedal buttons. Your footpedal may already support emulating a key combination. Please check
the footpedal settings or search the internet to see if it's supported. If not, then you can use a
program to do the emulation. Pedable is one such program for Windows. The following steps are
for Pedable.
Try the following steps on the practice files first.

 Close the Editor if already open.

 Download and install Pedable from http://www.nch.com.au/footpedal/index.html.
 Configure Pedable to send key combination for each footpedal button. Set the right
button to Ctrl+P.
 Close ExpressScribe if it's running. Close/minimize all other windows except your
Browser and Pedable.
 Press the right button on your footpedal. The print screen dialog should show if it's
working correctly.
 Open the Editor and set the key combination for the left pedal to Ctrl+O which is the
default for skip backwards/rewind.
Your footpedal should now be set up to use the right pedal for play/pause and left pedal for
rewind if the above steps were followed correctly.

Screening
Report issues correctly in order to receive more work.
We may ask you to report any audio issues in an assignment in order to determine the difficulty
level of the file. These reports are checked by our administrators who judge the difficulty level.
We charge our customers a higher rate for difficult files and also pay a higher rate to our
transcribers to compensate for the additional time and effort. Therefore it is of utmost importance
that any issues in a file are identified as early as possible. Our goal is to allow only good quality
files with American speakers into our system and screening helps us do that.
Our screening algorithm tracks your performance as a screener based on your reports and it may
affect the number of files available for transcription in your account at any given time. If you are
able to identify the issues correctly over a period of time, there will be more transcription files
available to you. Otherwise only the files which have been screened will be listed in your
available files tab.
We also provide a Cancel Warning for such files as the likelihood of the customer canceling the
order is high in case the file is found to be difficult. We recommend that you select other files for
transcription so as to avoid the loss of time and effort.

Best Practices
Based on our experience, we recommend the following to keep your account in good standing.

 Always preview the file before selecting. Check the audio quality and accents and select
only if you are comfortable with it. Always pick the easiest file.
 Do not select files which have audio issues or accented speakers. They are likely to be
returned to the customer.
 Don't use a footpedal. Use the keyboard shortcuts in the editor instead. It is far more
flexible and powerful and also very useful in reviews and proofreading.
 Use the automated transcript feature and correct it before submitting.
 Review your transcript once before submitting, at least for the first 10 files.
 As per your own assessment, if you cannot score a minimum of 3 on any assignment then
cancel it instead of submitting.
 Be extremely careful with the first 10 files. If you do not score good grades in those files,
transcription will be disabled and we will not be able to do anything about that.

Get Started
Please visit the Transcription page to start.